This thesis describes the method to recognition Korean noun phrase for parsing. Parsing means to analyze the role of the ingredients in a sentence and their correlation. Traditional parsers for systactic analysis have accuracy and robustness in only l...
This thesis describes the method to recognition Korean noun phrase for parsing. Parsing means to analyze the role of the ingredients in a sentence and their correlation. Traditional parsers for systactic analysis have accuracy and robustness in only limited domains. This is because natural languages have ambiguity in various domains and because complexity of general parsers increases proportionately with sentence length. To solve these difficulties, this thesis proposes noun phrase recognizer for syntactic analysis.
There are two methods to recognize noun phrase. One is rule-based approach and the other is statistical approach. The former can be high accuracy. But it is not easy to extend to new domain. Contrary to this, by the latter, it is not difficult to extent to new domain but accuracy is low relatively.
In this thesis, the noun phrase recognizer consist of the three parts: The one is to recognize noun phrase with information of syntactic characteristic. The second is to recognize noun phrase corresponded with temporal and location words. And the other is to recognize noun phrase with countable noun and its unit.
Additionally, this thesis considers the problem about compound noun as a pair of meaning and solves this by statistical-bases approach.
Experiment results shows high accuracy of Korean Noun Phrase Chunker using Korean information.
，免费韩语论文，韩语论文 | 
